Redes neurais de múltiplas camadas para redução do tempo de aquisição de dados para testes modais em estruturas flexíveis

Abstract

This work presents techniques to improve the dynamics parameters estimation of Single Degree of Freedom (SDOF) and Multiple Degree of Freedom (MDOF) models characteristic from flexible structures. The analyses are referred to the method that uses the Frequency Response Function (FRF) obtained from the impulse response of the flexible structure. We use for assumption that the considered models are convenient for a suitable description of the system. Thus, an experimental good method of obtaining the FRF should produce a significant accordance between the theoretical and the experimental FRF. The improvement in increasing the acquisition time artificially (forecasting) is analyzed by using a Multilayer Neural Network (MNN) model. The performance of neural forecaster is compared with results obtained using ARX and ARMAX models. The obtained results in this research, suggest the viability to use the MNN.